Displayr AI incorporates advances in AI technology into our software to make it easier for you to analyze your data. These methods are built off of OpenAI technology to provide new and improved features in the software. We understand that some customers may have strict data policies, so these enhanced Displayr AI features are available on an opt-in basis. You can opt-in or opt-out on an account-wide basis. Our current home-grown AI offerings remain available for those who decide not to opt into tools that utilize that specific technology.
This article covers:
Opting In or Out of Displayr AI
Enabling or disabling Displayr AI features is done on the company account-level, and can be done by Administrators and any user with editing privileges on an account. This field can be viewed and edited in Account Settings > General > Displayr AI. Displayr AI is not automatically enabled.
Before you opt in or out, whenever you try to use a Displayr AI feature directly or an improved Displayr AI feature is available, such as with automatic text categorization, you will be prompted to Accept (opt-in), Decline (opt-out) the Displayr AI - Terms of Use. Clicking Cancel will not run the feature and you will be prompted again about enabling next time.
Selecting Accept or Decline will automatically populate the Displayr AI field in Account Settings and note the User who made the change, as well as the date and time the change was made. Changes to this field are also tracked in the Account Settings > Usage > Audit Trail report.
Features
Better Auto Text Categorization
When Displayr AI is enabled, automatic categorization done within the Text Categorization tool will use the new, better categorization feature. The new feature is better at arriving at sensible categories as well as coding responses into the appropriate categories. The automatic categorization is done by default when selecting a new Text Categorization using the text categorization tool or via the Auto button in the Text Categorization tool itself. You will be notified whether or not Displayr AI is being used in the progress window that pops up:
If Displayr AI is not enabled, the categorization tool uses the original modeling to create categories as before. However, for both the initial categorization (where categories are created automatically when you open the tool) and the Auto button (where you can manually run the auto-categorization), you will now have the ability to specify the number of categories that are created - the default is 10:
You can see an example in the difference in results below based on a question "What are the main differences in cola drinkers?". The old categorization is on the left and you can see some categories are not useful/nonsensical whereas the Displayr AI categories are in the example on the right and much more relevant.
Better Variable Labeling on Import
Even with Displayr AI enabled, this feature is optional. Note that this is only an option when you initially import the data set and cannot be used when updating data sets previously imported. You can check the Use Displayr AI to tidy variable set names box at the bottom of the Where is your data window when importing a data set for the first time.
While Displayr is analyzing the file, you'll see this message. It does take longer than the simpler logic used in our regular import method:
During the import, Displayr will analyze the labeling of all of the variables in your data set and revise individual variable as well as variable set labeling to something more user-friendly. Otherwise, How Variable Sets/Questions are Automatically Grouped and Labeled provides more detail on how labeling has been done historically. After import, you should decide if you'd like to keep the Displayr AI labeling or use Displayr's original method of labeling, as Displayr AI labeling will be retained if you later update the data (see note below for more detail).
If you don't like the Displayr AI results, you can manually change the labels or delete and reimport the data without Displayr AI turned on to use the old method. Otherwise, the only way to get to the "original" labeling convention is to turn off Displayr AI via the Account Settings, and manually Reset > Variables from the toolbar (which will also break any sets apart) and Combine again. In certain circumstances, this can cause rework in the document.
The difference in labeling results is below. The left picture shows variables labeling based on our regular method and on the right is using Displayr AI:
Better Variable Set Labeling on Combine
When Displayr AI is enabled, when you Combine variables together into a variable set, the AI will automatically determine the variable set name. You can always Rename this afterwards, if need be. These labels will be retained if you later update the data (see note below for more detail). Previously, variable sets were labeled based on matching text across variables, see How Variable Sets/Questions are Automatically Grouped and Labeled for more detail. This You can see an example below. The variable set name q2 on the left was originally used due to the label pattern matching whereas the Displayr AI variable set name on the right is much more descriptive:
Technical details
Note on updating data
When Displayr AI is used for labeling sets on import as well as Combine, it is akin to manually changing the default labeling. That is, these labels will be retained when the data set updates. If instead, you want to ensure that your labeling is updated based on new data files, you should not use Displayr AI for labeling on import and you should disable it in your Account Settings so it does not relabel things when you Combine.
Nature of AI results
AI tools are in general not deterministic. As in, they don't give the same answer every time you run them. It's the nature of the tool itself, and if you have played around with ChatGPT you have probably experienced this yourself. Thus, while results are still smarter than what you can typically create via traditional machine learning alone, they are not perfect and sometimes can give a sub-par or nonsensical results. The implications of this are:
- If you receive poor results, try running the feature again to see if it improves.
- If there is an error or the feature does not work as expected, please take a screenshot of the issue and forward this along with a link to the document to support@displayr.com. Saving the result and noting the date/time will allow teams to more closely review the data to see what might have happened. So for example, if you get a bad outcome in a text categorization be sure to save the categories to the data set.
Feedback like this is valuable as it can help identify potential issues that need to be addressed by our developers.
Availability
Displayr AI is available on a regional basis to comply with region-specific data governance. It soon will be released to those using our Canadian and Australian servers. It will not available in our South East Asia region (southeastasia.displayr.com) due to a lack of the necessary infrastructure.
You can read about it in more detail in our Displayr AI - Terms of Use.